Pages:
Author

Topic: Sapphire R9 290X Tri-X OC Hashrate? - page 5. (Read 36569 times)

hero member
Activity: 700
Merit: 500
February 13, 2014, 09:30:20 AM
#24
Hi Guys,

I try to build a rig with 6 R9 290 Tri-x.

With windows or Linux, The computer just recognize 5 of them.
An idea to put the 6 on the same rig ?


ps : motherboard is Asrock probtc and my pci riser are on usb.



Thanks,
First step is to figure out what the problem spot is - card, cable, or mobo slot.  I would.definitely suggest linux for all testing.

1. Identify the failing card by running 100% fan and noting the one that doesn't spin up.
2. Swap pci-e ports, try again. If a different card fails, it is then pci-e slot. Tell us which one.
3. Swap riser (change both 1X, 16X, usb, and power). If the other card fails, you probably have a faulty riser cable.
4. Finally, if 2/3 had the same card fail consistently, plug it directly into the PCI-e 16x slot, and leave all other cards unplugged. Does it fail to boot? Faulty card

Faulty USB risers are fairly common. They also seem to burn out sometimes if you have too many plugged into a single SATA or Molex cable.
newbie
Activity: 25
Merit: 0
February 13, 2014, 03:30:36 AM
#23
Hi Guys,

I try to build a rig with 6 R9 290 Tri-x.

With windows or Linux, The computer just recognize 5 of them.
An idea to put the 6 on the same rig ?


ps : motherboard is Asrock probtc and my pci riser are on usb.



Thanks,
newbie
Activity: 31
Merit: 0
February 11, 2014, 12:08:58 PM
#22
I don't have 290X Tri-X... just Sapphire R9 290 reference. But here's some good settings, and perhaps a good pattern for them that could give good results for the 290X Tri-X Smiley

https://bitcointalksearch.org/topic/good-settings-for-r9-290-460181
newbie
Activity: 15
Merit: 0
February 09, 2014, 05:37:01 AM
#21
He said - 15% rejects - is this Hrate worth this?
newbie
Activity: 23
Merit: 0
February 09, 2014, 04:52:58 AM
#20
running 2x 290X.
Gets between 1170-1190k/ each
temp average between 73-75


What settings are you using?
Sgminer, karloth or cgminer 3.7.2?

I never got more than 970 from my Sapphire R2 290X
newbie
Activity: 15
Merit: 0
February 08, 2014, 09:49:49 PM
#19
Not exactly uncommon - you could RMA it, but is it worth a 10% hashrate boost to lose time waiting for a replacement?  Have you tried 947 engine?

I found these card are undervolted - when high stressed it just cant make it - but in games - ok , yet...

orginal vddc - 1.020 (stress) at 1.070 (stress) should be fine still testing

I have to check every single one
sr. member
Activity: 473
Merit: 250
"Proof-of-Asset Protocol"
February 08, 2014, 12:20:17 PM
#18
running 2x 290X.
Gets between 1170-1190k/ each
temp average between 73-75


Outside temp: -1 to 2
Fan in Auto Mode.

but notice 15% invalid shares.
sr. member
Activity: 336
Merit: 254
CoinMine.pw
February 08, 2014, 12:04:47 PM
#17
This said, my 945KH/s+ cards run at a cool 65 degrees C max (lowest is 57 degrees).
Stock unmodified reference cards.  899KH/s stable as a rock too.
You set 100% fan. And ambient temperature 10`C? Smiley
I have tried 100% fan its quite loud and got 69`C. Ambient temperature 21`C.
Set back to 55% fan (2740 RPM). It's 74`C back but quite acceptable.
sr. member
Activity: 473
Merit: 250
"Proof-of-Asset Protocol"
February 08, 2014, 03:20:55 AM
#16
running 2x 290X.
Gets between 1170-1190k/ each
temp average between 73-75
hero member
Activity: 700
Merit: 500
February 07, 2014, 02:50:39 PM
#15
whats your vcore's for these cards?
Stock.  I never use windows, so can't touch the voltages on Hawaii chips yet.

got 5 pieces.

one is going dead on stock (gpu0) trying on lower gpu speed (now 900 , stock 1000 ) at 1000 is going dead after few seconds - probably I dl RMA.
Mem clock lowered 1300->1250


cooling is same for all - but temps are weird high.
Seems like they had been assembled in rush!
Not exactly uncommon - you could RMA it, but is it worth a 10% hashrate boost to lose time waiting for a replacement?  Have you tried 947 engine?

You can re-thermal paste the cards if you don't mind losing your warranty, and it might help even things out while shedding a few degrees.  I have also observed some cards that read at higher temperatures but don't actually seem to be (e.g. the measurement seems skewed).  As long as they are within range, I wouldn't be too concerned.

I do agree they seem like they were assembled in a bit of a rush.  But not as bad as the new MSI 290 cards... those are bloody weird, but good hashers anyway.
newbie
Activity: 15
Merit: 0
February 07, 2014, 04:43:42 AM
#14
whats your vcore's for these cards?
got 5 pieces.

one is going dead on stock (gpu0) trying on lower gpu speed (now 900 , stock 1000 ) at 1000 is going dead after few seconds - probably I dl RMA.
Mem clock lowered 1300->1250


cooling is same for all - but temps are weird high.
Seems like they had been assembled in rush!

http://i61.tinypic.com/wj80h3.jpg
hero member
Activity: 700
Merit: 500
February 07, 2014, 12:30:19 AM
#13
This said, my 945KH/s+ cards run at a cool 65 degrees C max (lowest is 57 degrees).

Looks like you don't use standard Sapphire Tri-X card. It's quite hot. 74'C is minimum for me Sad

Of course I use standard Sapphire Tri-X cards...  Sapphire R9 290 Tri-X OC, no mods.  I just have plenty of experience running rigs.  

I have banks of reference Sapphire R9 290s running at <60 degrees too:


Stock unmodified reference cards.  899KH/s stable as a rock too.
sr. member
Activity: 336
Merit: 254
CoinMine.pw
February 06, 2014, 11:37:49 PM
#12
This said, my 945KH/s+ cards run at a cool 65 degrees C max (lowest is 57 degrees).

Looks like you don't use standard Sapphire Tri-X card. It's quite hot. 74'C is minimum for me Sad
hero member
Activity: 700
Merit: 500
February 06, 2014, 10:35:56 PM
#11
It's better to check Watt rate and temperature also.
Why?  Power consumption is almost totally irrelevant at current scrypt profitability.  You could use double the power for 10% more hashrate and probably come out ahead right now.

This said, my 945KH/s+ cards run at a cool 65 degrees C max (lowest is 57 degrees).

PS: I use TC: 20481 don't know where I found it, but it give me better rate, than higher values

Would be shaders * 8 + 1, which makes sense.  I find shaders * 10 + 1 to be solid on these cards, but this also won't make a huge difference.


PS: xI don't make much difference. I get almost same rate from 200 to 600
If by "almost the same rate" you mean several percent, sure.  But yes, the key is really that using xI over I makes a decent difference.
sr. member
Activity: 336
Merit: 254
CoinMine.pw
February 06, 2014, 06:31:34 PM
#10
It's better to check Watt rate and temperature also.

A have 2 of 290(not X) Tri-X and get:

863 KH/s each with 632 Watt and temperature 77`C (965KHz Core, 1250 Ram)
905 KH/s each with 677 Watt and temperature 88`C (1000KHz Core, 1500 Ram)

It's 4.8% rate gain and 7.1% Watt gain. Also I don't like 88`C and possible card crash.

So it mine already 6 days 24/7 without problems. I use sgminer 4.0.0  xI:400

PS: I use TC: 20481 don't know where I found it, but it give me better rate, than higher values
PS: xI don't make much difference. I get almost same rate from 200 to 600
hero member
Activity: 700
Merit: 500
February 06, 2014, 02:09:10 PM
#9
I've now got it set up with a .conf file and running kalroth's 3.7.2 cgminer and can't get as high of a hash as I was getting before without the xintensity.  I think my card has 2816 shaders so the xintensity would be 372?

Try bumping it up a good chunk, say double.  Personally I didn't have much luck with kalroth's miner, even though it has the same xintensity code as sgminer.  372 makes sense by my math.  I use 1100 on my 290 (non-X).

Also set your tc to a multiple of 2816, and then add or subtract 1 to see which gives more efficiency (usually adding 1 will give you a bump).  Say, 28161 or something.  xintensity means you don't have to push TC so high to get the same performance, because you are aligning the number of threads launched with the number of threads you can actually run at one time.
newbie
Activity: 34
Merit: 0
February 04, 2014, 04:51:32 PM
#8
I've now got it set up with a .conf file and running kalroth's 3.7.2 cgminer and can't get as high of a hash as I was getting before without the xintensity.  I think my card has 2816 shaders so the xintensity would be 372?
hero member
Activity: 700
Merit: 500
February 04, 2014, 03:20:06 PM
#7
I've tried about a dozen other higher GPU engine values higher and they all resulted in a lower hash by about 100-150Kh/s lower.
What GPU engine value do you recommend I try? I've never heard of xintensity, I will definitely look into it.  As a side note. I've never noticed any difference at all in performance with the following included:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
Am I doing something wrong with the above code?
You only need to run the setx command _once_ on a windows machine, and the variable will be fixed forever.  If you don't do GPU_MAX_ALLOC_PERCENT, cgminer simply won't work with higher threadconcurrencies.

GPU engine depends massively on BIOS.  Some 290 (and X) BIOSes will drop in performance past 1000, others drop past 1020, and still others keep climbing all the way up to and beyond 1080.  Typically 1000 seems to work on everything tho, if it doesn't, try another BIOS (e.g. a different manufacture like Sapphire).

My favorite BIOSes on the 290 at the moment are:
Powercolor AXR9 290 4GBD5-MDH/O  015.042.000.000.000000  (does 930KH/s at 1499/1040 mem/engine)
Sapphire R9 290 4096 MB 015.041.000.000.000000 (does 899KH/s at 1499/1000 mem/engine - DO NOT USE THE 015.041.000.001.003747 BIOS, you will lose 20-30KH/s at the same clocks)
and
Sapphire R9 290 4096 MB Tri-X OC 015.042.000.000.000000 (does 961KH/s at 1499/1080 and 954KH/s at 1499/1070 mem/engine)

You will have to experiment a bit to find the best BIOS for your 290X.
newbie
Activity: 34
Merit: 0
February 04, 2014, 02:52:09 PM
#6
I've tried about a dozen other higher GPU engine values higher and they all resulted in a lower hash by about 100-150Kh/s lower.
What GPU engine value do you recommend I try? I've never heard of xintensity, I will definitely look into it.  As a side note. I've never noticed any difference at all in performance with the following included:
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_USE_SYNC_OBJECTS 1
Am I doing something wrong with the above code?
hero member
Activity: 700
Merit: 500
February 04, 2014, 02:03:38 PM
#5
My Gigabyte R9 290X is getting 863Kh/s with WU:780/m and 79C

setx GPU_MAX_ALLOC_PERCENT 100
cgminer --scrypt -o stratum+tcp://scrypt.megamultipool.com:80 -u user.user -p password --intensity 20 -g 1 -w 256 --lookup-gap 2 --thread-concurrency 24550 --gpu-engine 920 --auto-fan --gpu-memclock 1500 --gpu-powertune 20 --temp-target 80 --temp-cutoff 95 --temp-overheat 92

Why do you have GPU engine so low?  And stop using intensity - switch to xintensity and you will see several percent hashrate increase.
Pages:
Jump to: